Hybrid Auto Text Summarization Using Deep Neural Networks And Fuzzy Logic System

Hybrid Auto Text Summarization Using Deep Neural Networks and Fuzzy Logic System

As the amount of data available for a particular topic is increasing day by day due to which the availability of bluk is also increased ,out of which some my be relevant to users search while some may be irrelevant. In such case to find relevant data for users search and to save time it is necessary to have small summary of the text document. Summary generated by users is time consuming and tedious . So there is a need for automatic summarizer Which condense text document into few lines of summary.

Four Feature are used:Title similarity:- Sentences is considered to be important if it matches the title of the document. Term weight:-Sentence is important if the word it contains appears frequently in the text document. Named entities:- Sentence is important if it is a proper noun , since named entities usually contain key Information. Numerical data:-Sentence is important if it contains numerical values , statistics etc. The features are applied to sentences based on priority of The feature considering the document typea)If the document contains more numerical data the high priority is given to the numerical data feature. b)If the whole document contains only words the priority is given to tiltle similarity , term weight and named entity and so on. Based on these features sentence matrix is formed.

Three rule base applied are:a)If the scores are high then the sentence is important. b)If the scores are medium then the sentence is average. c)If the scores are low then the sentence is unimportant.

Preprocessing:a)Stop word removal:-Here the stop words such as “a”, ”an”, ”the” along with special characters are considered as unimportant and removed. b)Stemming the word is bought to its root by removing ing, ed etc from the word. Sentences having high rank are selected and top sentences with highest is selected and arranged according to their apperances in the document